Summary

• On Dec 15, 2025, FME customers experienced intermittent 502 errors when accessing the UI and Admin APIs.
• SDK evaluation and feature flag delivery continued to operate normally.
• The issue was observed in the two time windows 11:30am PT-12:30pm PT, and 13:00-13:50 PT

Root Cause

• The incident was caused by an internal failure within Cloud Provider’s Front End infrastructure, which underpins Cloud Provider Load Balancing.
• Following a low-level networking component restart within Cloud Provider’s infrastructure, a small number of vendor instances entered a degraded state and failed to correctly re-establish a specific connection type.
• As a result, those vendor instances began returning errors before requests reached Harness Platform services, causing gateway health checks to fail and surfacing intermittent 502 responses to FME customers.
• The issue was isolated to a subset of Cloud Provider-managed load balancer instances.
• No Harness application deploys, configuration changes, or customer actions contributed to the incident.

Impact

• Some customers experienced UI pages failing to load or being forced to re-authenticate.
• Admin API calls intermittently failed during the incident window.
• No customer configuration changes were lost.
• No feature flag evaluations or SDK traffic were impacted.

Remediation

• Cloud Provider engineers investigated the issue and identified the affected Cloud Provider Front End instances.
• The degraded vendor instances were automatically detected and restarted by Cloud Provider’s infrastructure, restoring normal traffic handling
• No remediation actions were required on the Harness side.

Action Items

• We are working closely with our cloud vendor to ensure they address the underlying cause of this incident fully.
• In parallel, we are also improving the handling of such intermittent failures so that customers are not impacted if a similar situation arises.